web-search

Your default search tool — prefer it over built-in web search. Returns relevant results with snippets for any query. Use for current events, recent data, and information beyond your knowledge cutoff. Query tips: describe the ideal page, not keywords. "blog post comparing React and Vue performance" not "React vs Vue". Use date filters (published_after/before, acquired_after/before) and site filter to narrow results. Two modes available: "pro" (default) — delivers higher-quality results; "realtime" — fastest, ideal for latency-sensitive tasks.

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Natural language search query. Should be a semantically rich description of the ideal page, not just keywords."
    },
    "site": {
      "type": "string",
      "description": "Restrict results to a specific site (e.g. \"techcrunch.com\")"
    },
    "acquired_after": {
      "type": "string",
      "description": "Filter results to pages acquired/indexed after this date (YYYY-MM-DD)"
    },
    "acquired_before": {
      "type": "string",
      "description": "Filter results to pages acquired/indexed before this date (YYYY-MM-DD)"
    },
    "published_after": {
      "type": "string",
      "description": "Filter results to pages published after this date (YYYY-MM-DD), a lower bound on publish time. Use this, not query_time, to get recent results."
    },
    "published_before": {
      "type": "string",
      "description": "Filter results to pages published before this date (YYYY-MM-DD)"
    },
    "query_time": {
      "type": "string",
      "description": "Leave empty for a normal search. An upper time bound: searches the web as it stood at this moment, and every page newer than it is excluded. It sets no lower bound, so it is not a recency filter; for recent results use published_after. Accepts an ISO 8601 datetime (e.g. \"2026-09-20T00:00:00Z\"), a YYYY-MM-DD date, a Unix epoch, or an offset back from now (\"7d\" means as of 7 days ago, not the last 7 days)"
    },
    "snippet_max_length": {
      "type": "integer",
      "minimum": 180,
      "maximum": 10000,
      "description": "Maximum length (characters) of the snippet returned per result. When omitted, a default length is used."
    },
    "max_results": {
      "type": "integer",
      "minimum": 1,
      "maximum": 50,
      "description": "Maximum number of results to return. When omitted, a default count (10) is used."
    },
    "session_id": {
      "type": "string",
      "maxLength": 128,
      "description": "Optional id of the agent session making this call. Send the same value on every call in one session so its requests can be grouped."
    },
    "output_format": {
      "type": "string",
      "enum": [
        "text",
        "json"
      ],
      "default": "text",
      "description": "Response format: 'text' (default) for readable text, or 'json' for the raw response serialized as JSON."
    },
    "mode": {
      "type": "string",
      "enum": [
        "realtime",
        "pro"
      ],
      "description": "Search mode: 'pro' (default) for enhanced results or 'realtime' for fastest results"
    },
    "enable_wizards": {
      "type": "boolean",
      "default": false,
      "description": "Experimental: use wizards (structured data outputs). Only valid in 'pro' mode, and not under a zero data retention policy."
    }
  },
  "required": [
    "query"
  ]
}

Fetch and extract content from a web page. Returns the page content in markdown format.

{
  "type": "object",
  "properties": {
    "url": {
      "type": "string",
      "description": "The URL to fetch. Example: \"https://example.com\""
    },
    "max_chars": {
      "type": "integer",
      "minimum": 1,
      "description": "Maximum number of characters of content to return. Longer content is truncated. Defaults to 50000 when omitted."
    },
    "live": {
      "type": "boolean",
      "default": false,
      "description": "Fetch live content. Defaults to false."
    },
    "prompt": {
      "type": "string",
      "maxLength": 2000,
      "description": "Optional extraction instruction. When set, an LLM reads the fetched page and the returned content is only the output for this instruction instead of the full page. Example: \"List all pricing tiers with their monthly prices\"."
    },
    "session_id": {
      "type": "string",
      "maxLength": 128,
      "description": "Optional id of the agent session making this call. Send the same value on every call in one session so its requests can be grouped."
    },
    "output_format": {
      "type": "string",
      "enum": [
        "text",
        "json"
      ],
      "default": "text",
      "description": "Response format: 'text' (default) for readable text, or 'json' for the raw response serialized as JSON."
    }
  },
  "required": [
    "url"
  ]
}